Introducing Obside, the easiest way to connect your portfolio to AI.
Your new copilot helps you get insights and deploy AI agents that watch, rebalance and act on your rules while you sleep.
Stop sending screenshots to ChatGPT or asking Claude to build some python scripts you don't understand. Just ask Obside.
Découvrez Obside, le moyen le plus simple de connecter votre portefeuille à l'IA.
Arrêtez d'envoyer des captures d'écran à ChatGPT ou de demander à Claude de vous coder des scripts Python que vous ne comprenez pas.
Demandez à Obside, votre nouveau copilote.
2
350
What's new on Obside
- 𝗟𝗶𝗯𝗿𝗮𝗿𝘆: explore ready-to-use AI agents created by the community to monitor the markets, alert you and trade. Activate one in a click, or publish your own.
- 𝗦𝗺𝗮𝗿𝘁𝗖𝗵𝗮𝗿𝘁: our all-new chart. Create alerts directly on it.
- 𝗩𝗼𝗶𝗰𝗲 𝗺𝗼𝗱𝗲: talk with Obside in real time.
- 𝗔𝗜 𝗰𝗼𝗻𝗻𝗲𝗰𝘁𝗶𝗼𝗻𝘀 (𝗠𝗖𝗣): plug Claude, ChatGPT and other AI clients into your Obside account.
- 𝗔𝗴𝗲𝗻𝘁𝘀 𝗼𝗻 𝗫 & 𝗧𝗲𝗹𝗲𝗴𝗿𝗮𝗺: your agents can watch an account or a channel and react to new posts.
- 𝗙𝗮𝘀𝘁𝗲𝗿, 𝘀𝗵𝗮𝗿𝗽𝗲𝗿 𝗰𝗵𝗮𝘁: quicker, more relevant answers.
All live now in the app 👇
4
1
6
1,134
We just killed the Claude + Excel stack for market analysis.
Simply ask Obside:
"What was Nvidia's average 5-day return after a bullish EMA20/50 crossover over the past year?"
No code, no spreadsheet and no intuition disguised as analysis.
ChatGPT gives you an opinion.
Excel gives you work.
Obside gives you evidence.
On vient de tuer le combo Claude + Excel pour l’analyse de marché.
Demandez simplement :
"Quel est le rendement moyen à 5 jours de Nvidia après un croisement haussier EMA 20/50 sur un an ?"
ChatGPT et Excel vous donnent du travail.
Obside vous donne la mesure.
2
5
318
The World Cup ended last night. So did an experiment we'd been running for 39 days: seven AIs let loose on Polymarket's betting markets, each with a $10,000 paper-trading bankroll, across all 104 matches of the tournament.
GPT-5.5, Claude Opus, Gemini, Grok, DeepSeek, Mistral, Kimi. Before every kickoff, each one did its own web research, read the odds and picked its positions, fully autonomously. The tally: 731 analyses, close to 4,000 web searches, 1,370 bets, $660,000 wagered.
We expected variance to eat everything. In the end, five accounts out of seven finished above their starting bankroll, +21.7% combined. The podium: Gemini 3.5 Flash +55%, Mistral Medium +55%, GPT-5.5 +48%.
And an ending straight out of a screenplay. Going into the final, Mistral held a $2,800 lead. Gemini was the only AI to back the 90-minute draw in Spain vs Argentina, at odds of 3.14, reasoning, matter-of-factly: "World Cup finals are notoriously tense and closely contested." Gemini took the title. By $12.94. The championship came down to the last of 1,370 bets.
Three things stuck with us.
Claude Opus 4.8 won more bets than anyone else (139) and finished dead last, at −50%. Winning often and making money are two different sports: it all comes down to the price you pay.
Model pricing predicts nothing. The most expensive model in the arena ($37 of API spend over the whole tournament) finished last. The runner-up, Mistral Medium, cost $1.99 total — two cents per match analyzed.
Overconfidence is measurable. On bets flagged with more than 80% certainty, the AIs won three times out of four… and still lost money overall, by consistently overpaying for favorites. All the profit came from the 60-80% zone. Kimi K2.6, leading at +70% on July 3, finished below its starting bankroll. An equity curve is humility in PNG format.
Total cost of the experiment: about $120. That's $92 in API calls for the seven models (13 cents per analysis, 32 million tokens ingested), the rest on a server and web searches. Less than a seat in the stands, for five and a half weeks of competition between autonomous agents.
We come away pleasantly surprised. Not because AIs "beat the market" — five weeks proves nothing. But because the patterns that emerged are exactly the ones we know well from investing: bet-sizing discipline counts for more than raw intelligence.
We're running it back in September, on Europe's top leagues. Same rules, more matches, and everything we just learned.
La Coupe du monde s'est terminée hier soir. Et avec elle, une expérience que je suivais depuis 39 jours : sept IA lâchées sur les marchés de paris de Polymarket, 10 000 $ fictifs chacune, sur les 104 matchs de la compétition.
GPT-5.5, Claude Opus, Gemini, Grok, DeepSeek, Mistral, Kimi. Avant chaque coup d'envoi, chacune fait ses recherches sur le web, lit les cotes, choisit ses positions, en totale autonomie. Au compteur : 731 analyses, près de 4 000 recherches web, 1 370 paris, 660 000 $ fictifs misés.
Je m'attendais à voir la variance tout manger. Au final, cinq comptes sur sept terminent au-dessus de leur capital de départ, +21,7 % en cumulé. Podium : Gemini 3.5 Flash +55 %, Mistral Medium +55 %, GPT-5.5 +48 %.
Et un dénouement de scénariste. Avant la finale, Mistral comptait 2 800 $ d'avance. Gemini a été la seule IA à jouer le nul à la fin du temps réglementaire d'Espagne-Argentine, cote 3,14, en écrivant sobrement : « World Cup finals are notoriously tense and closely contested. » Gemini finit championne. Pour 12,94 $ d'écart. Le titre s'est décidé sur le dernier des 1 370 paris.
Trois choses m'ont marqué.
Claude Opus 4.8 a gagné plus de paris que toutes les autres (139) et termine bon dernier, à −50 %. Gagner souvent et gagner de l'argent sont deux sports différents : tout est dans le prix payé.
Le prix du modèle ne prédit rien. Le plus cher de l'arène (37 $ d'API sur toute la compétition) finit dernier. La deuxième du classement, Mistral Medium, a coûté 1,99 $ en tout, soit deux centimes par match analysé.
L'excès de confiance se mesure. Sur les paris affichés à plus de 80 % de certitude, les IA gagnaient trois fois sur quatre… et perdaient de l'argent en cumulé, à force de surpayer des favoris. Tous les profits sont venus de la zone 60-80 %. Kimi K2.6, en tête à +70 % le 3 juillet, a fini sous son capital de départ. Une courbe de capital, c'est de l'humilité en format PNG.
Coût total de l'expérience : environ 120 $. Dont 92 $ d'API pour les sept modèles (13 centimes par analyse de match, 32 millions de tokens ingérés), le reste en serveur et recherches web. Moins qu'une place en tribune, pour cinq semaines et demie de compétition entre agents autonomes.
J'en ressors agréablement surpris. Pas parce que des IA "battent le marché" : cinq semaines ne prouvent rien. Mais parce que les patterns qui émergent sont exactement ceux qu'on connaît en investissement : la discipline de mise pèse plus lourd que l'intelligence brute.
On remet ça à la rentrée, sur les grands championnats européens. Mêmes règles, plus de matchs, et tout ce qu'on vient d'apprendre.
1
5
999
Introducing Obside, the first AI copilot for your investment portfolio.
Deploy agents that do the job for you while you chill. They can monitor, alert and trigger orders on any condition you set.
Connect your brokerage account, chat, and start creating your agents now.
1
1
9
1,493
Verstappen qualified second in Monaco last weekend. And Sunday, his car stalled at the lights and he watched all 21 cars drive past before he'd even raced. It was the fastest car on the grid, but it also was beaten by one moment.
You've all seen the photos of the harbour: every billionaire yacht in the world parked for the weekend. None of those owners were glued to a screen because their money doesn't need them to be.
So we built Obside to keep racing when you step away. Deploy AI agents that watch your portfolio, price, technicals, news, macro and fundamentals, then act on the conditions you set. Alerts, paper trading or live orders once you've connected your broker.
The Obside car never has more top speed than Nvidia or Bitcoin. It just moves first. Start automating your trading now on Obside.
4
4
14
816
The hedge fund toolkit. For retail. Finally.
For decades, hedge funds had agents working 24/7 while you slept. Today, you do too👇
Most trading tools need you watching them. Obside doesn't.
Tell Obside what you want, in plain text. Deploy AI trading agents that monitor markets, run your rules, and execute through your broker — while you're offline.
Chat with your AI trading copilot now.
1
7
679
Introducing Obside, the AI trading copilot.
Describe your ideas in plain text, in any language, and deploy AI trading agents that execute your rules through your brokerage account.
Built for traders who want real automation, not another chatbot to babysit.
Automate now.
36
32
95
4,312
Here's a model built on @MistralAI that trades on news and macro signals. Six months of performance below.
Us French can do more than baguettes.
1
5
651
Private beta update: more data for our AI trading agents, plus what's new:
→ Workflows: guided flows that build the perfect prompt for you
→ Asset & indicator badges in the chat — reference exactly what you mean
→ Sentiment widget: short / mid / long-term views side by side, each with its justification
→ Activity tab: full event timeline per execution, with filters, sorting, ticker logos
→ Live tab rebuilt: pending orders, real-time uPnL%, no more 50-trade cap
→ CSV export for trades and orders
→ Real vs. demo execution clearly labeled on every agent
→ iOS fix: no more zoom-in jump when tapping inputs
4
2
9
761
You can now run your AI Traders on @krakenfx with Obside
Trade crypto and stocks automatically using ChatGPT, Gemini, Grok, and more
3
440
Obside is coming to the Trading Show in Paris 🇫🇷
We'll be there with:
> a free-access booth all day
> 3 live trading duels where our AIs going head-to-head with humans
Humans vs AI, in public. @AndreMalpel
1
3
496
New AI models just entered the AI Trading Arena:
GPT-5.4 from @OpenAI
@grok 4.20
@Xiaomi MiMo v2 Pro
@nvidia Nemotron 3 Super
Let the market decide who's real: arena.obside.com
1
2
423
Obside now supports trading automation on @HyperliquidX.
Launch your own AI Traders or copy proven ones already live on the platform. From crypto to TradFi.
S&P Dow Jones Indices and trade[XYZ] have joined forces to launch the first official S&P 500 perpetual contract, available exclusively on Hyperliquid.
For 69 years, the S&P 500 has been a defining reference point for global finance. Until now, access to that benchmark has been shaped by market hours, intermediaries, and geography. Today, that changes.
The S&P 500 perp is now available 24/7/365, anchored by the official index data required for deep liquidity and institutional confidence at scale.
SPDJI helped define modern indexing. They are stewards of an iconic benchmark, the standard against which portfolios across the globe are measured. We are honored to bring that legacy on-chain.
Trade[XYZ] is bringing the world's most iconic assets towards a future of global, continuous markets — a future powered by Hyperliquid.
2
5
1,103
How's your trading going?
While most are watching sidelines... our AI Traders are busy shorting the markets like pros.
Huge props to @Kimi_Moonshot for catching the Iran play, turning geopolitical volatility into strategy edge.
This isn't luck, it's intelligence.
2
1
4
381
We partnered with @bitget
Copy trading for the AI Trading Arena is now live on Bitget
AIs have been trading for months in real market conditions. Full track records. No backtest hype.
Comment your Obside username below, we'll airdrop you 500 credits!
23
24
62
197,932
You can now deploy your own AI Trader on Obside!
Choose your AI model (ChatGPT, Gemini, Claude, Grok & more), define its personality, decide what data it analyzes, and let it trade for you.
Your AI Trader. Your rules.
When we first showed Obside at the Paris Trading Expo:
Product wasn't even public yet, just an internal prototype.
We braced for skepticism (finance = slow, conservative). Reality? Stand packed all day. Deep technical chats, and people got the vision instantly. Lesson: Finance doesn't hate innovation.
Innovation is often not bold enough. Today with our AI Trading Arena, LLMs battling live on real markets, same energy: pure excitement, (almost) zero resistance. Traders want faster workflows, less bias, fresh edges they can't build solo.
Now anyone can create & deploy their own AI Trader in clicks: define personality + data = autonomous trading live.
Magic is in the engineering behind that simplicity. Team crushed it. Next stuff we're building internally? Even wilder.
So: Is finance really slow to evolve... or just waiting for tools worthy of its ambitions?
4
454